NEZHA: A Zero-sacrifice and Hyperspeed Decoding Architecture for Generative Recommendations

• URL: http://arxiv.org/abs/2511.18793v1
• Date: Mon, 24 Nov 2025 05:53:46 GMT
• Title: NEZHA: A Zero-sacrifice and Hyperspeed Decoding Architecture for Generative Recommendations
• Abstract summary: NEZHA is a novel architecture that achieves hyperspeed decoding for Generative Recommendation (GR) systems without sacrificing recommendation quality.<n>We demonstrate the effectiveness of NEZHA through extensive experiments on public datasets and have successfully deployed the system on Taobao since October 2025.
• Score: 44.6848620438905
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Generative Recommendation (GR), powered by Large Language Models (LLMs), represents a promising new paradigm for industrial recommender systems. However, their practical application is severely hindered by high inference latency, which makes them infeasible for high-throughput, real-time services and limits their overall business impact. While Speculative Decoding (SD) has been proposed to accelerate the autoregressive generation process, existing implementations introduce new bottlenecks: they typically require separate draft models and model-based verifiers, requiring additional training and increasing the latency overhead. In this paper, we address these challenges with NEZHA, a novel architecture that achieves hyperspeed decoding for GR systems without sacrificing recommendation quality. Specifically, NEZHA integrates a nimble autoregressive draft head directly into the primary model, enabling efficient self-drafting. This design, combined with a specialized input prompt structure, preserves the integrity of sequence-to-sequence generation. Furthermore, to tackle the critical problem of hallucination, a major source of performance degradation, we introduce an efficient, model-free verifier based on a hash set. We demonstrate the effectiveness of NEZHA through extensive experiments on public datasets and have successfully deployed the system on Taobao since October 2025, driving the billion-level advertising revenue and serving hundreds of millions of daily active users.

Related papers

• HGenPush: A Heterogeneous Generative Recommendation Architecture for Industrial Push Notification Systems [19.17988315151231]
  We propose an end-to-end heterogeneous generative recommendation architecture called HGenPush.<n>HGenPush simultaneously fulfills users' demands for high-quality content and trusted authors.<n>We have deployed HGenPush on the push notification system of Kuaishou, a large-scale short-video platform, achieving a significant 0.181% increase in daily active users.
  arXiv  Detail & Related papers  (2026-07-03T14:18:19Z)
• AlphaGRPO: Unlocking Self-Reflective Multimodal Generation in UMMs via Decompositional Verifiable Reward [52.7128440950672]
  We propose AlphaGRPO to enhance multimodal generation capabilities without an additional cold-start stage.<n>Our approach unlocks the model's intrinsic potential to perform advanced reasoning tasks.<n>To address the challenge of providing stable supervision for real-world multimodal generation, we introduce the Decompositional Verifiable Reward (DVReward)
  arXiv  Detail & Related papers  (2026-05-12T17:59:47Z)
• Bending the Scaling Law Curve in Large-Scale Recommendation Systems [25.722146264492096]
  We present ULTRA-HSTU, a novel sequential recommendation model developed through end-to-end model and system co-design.<n>Our solution is fully deployed at scale, serving billions of users daily and driving significant 4% to 8% consumption and engagement improvements in real-world production environments.
  arXiv  Detail & Related papers  (2026-02-19T01:15:17Z)
• Self-Evolving Recommendation System: End-To-End Autonomous Model Optimization With LLM Agents [18.707716142982992]
  We propose a self-evolving system to autonomously generate, train, and deploy complex model changes.<n>Our agents act as specialized Machine Learning Engineers (MLEs)<n>The effectiveness of this approach is demonstrated through several successful production launches at YouTube.
  arXiv  Detail & Related papers  (2026-02-10T19:16:52Z)
• PRISM: Parametrically Refactoring Inference for Speculative Sampling Draft Models [7.294980307522086]
  Large Language Models (LLMs), constrained by their auto-regressive nature, suffer from slow decoding.<n>We propose PRISM, which disaggregates the computation of each predictive step across different parameter sets.<n>We show that PRISM boosts the decoding throughput of an already highly optimized inference engine by more than 2.6x.
  arXiv  Detail & Related papers  (2026-02-02T07:46:03Z)
• Noise Hypernetworks: Amortizing Test-Time Compute in Diffusion Models [57.49136894315871]
  New paradigm of test-time scaling has yielded remarkable breakthroughs in reasoning models and generative vision models.<n>We propose one solution to the problem of integrating test-time scaling knowledge into a model during post-training.<n>We replace reward guided test-time noise optimization in diffusion models with a Noise Hypernetwork that modulates initial input noise.
  arXiv  Detail & Related papers  (2025-08-13T17:33:37Z)
• VeriReason: Reinforcement Learning with Testbench Feedback for Reasoning-Enhanced Verilog Generation [9.07044866283158]
  We introduce VeriReason, a framework integrating supervised fine-tuning with Guided Reward Proximal Optimization (GRPO) reinforcement learning for RTL generation.<n>On the VerilogEval Benchmark, VeriReason delivers 83.1% functional correctness, substantially outperforming both comparable-sized models and much larger commercial systems like GPT-4 Turbo.<n>VeriReason represents the first system to successfully integrate explicit reasoning capabilities with reinforcement learning for Verilog generation, establishing a new state-of-the-art for automated RTL synthesis.
  arXiv  Detail & Related papers  (2025-05-17T05:25:01Z)
• AKD : Adversarial Knowledge Distillation For Large Language Models Alignment on Coding tasks [4.757470449749877]
  This paper introduces Adversarial Knowledge Distillation (AKD) to distill the capabilities of larger models into smaller, more efficient ones.<n>AKD provides a framework for enhancing model robustness, reliability, and security while improving their parameter-efficiency.
  arXiv  Detail & Related papers  (2025-05-05T22:41:19Z)
• Preventing Non-intrusive Load Monitoring Privacy Invasion: A Precise Adversarial Attack Scheme for Networked Smart Meters [99.90150979732641]
  We propose an innovative scheme based on adversarial attack in this paper.<n>The scheme effectively prevents NILM models from violating appliance-level privacy, while also ensuring accurate billing calculation for users.<n>Our solutions exhibit transferability, making the generated perturbation signal from one target model applicable to other diverse NILM models.
  arXiv  Detail & Related papers  (2024-12-22T07:06:46Z)
• HUWSOD: Holistic Self-training for Unified Weakly Supervised Object Detection [66.42229859018775]
  We introduce a unified, high-capacity weakly supervised object detection (WSOD) network called HUWSOD. HUWSOD incorporates a self-supervised proposal generator and an autoencoder proposal generator with a multi-rate re-supervised pyramid to replace traditional object proposals. Our findings indicate that randomly boxes, although significantly different from well-designed offline object proposals, are effective for WSOD training.
  arXiv  Detail & Related papers  (2024-06-27T17:59:49Z)
• Integrated Modeling, Verification, and Code Generation for Unmanned Aerial Systems [10.292890852621346]
  Unmanned Aerial Systems (UAS) are widely used in safety-critical fields such as industrial production, military operations, and disaster relief. This paper aims to investigate an integrated approach to modeling, verification, and code generation for UAS.
  arXiv  Detail & Related papers  (2024-06-13T14:53:40Z)
• RigorLLM: Resilient Guardrails for Large Language Models against Undesired Content [62.685566387625975]
  Current mitigation strategies, while effective, are not resilient under adversarial attacks. This paper introduces Resilient Guardrails for Large Language Models (RigorLLM), a novel framework designed to efficiently moderate harmful and unsafe inputs.
  arXiv  Detail & Related papers  (2024-03-19T07:25:02Z)
• RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment [32.752633250862694]
  Generative foundation models are susceptible to implicit biases that can arise from extensive unsupervised training data. We introduce a new framework, Reward rAnked FineTuning, designed to align generative models effectively.
  arXiv  Detail & Related papers  (2023-04-13T18:22:40Z)
• Dynamic Model Pruning with Feedback [64.019079257231]
  We propose a novel model compression method that generates a sparse trained model without additional overhead. We evaluate our method on CIFAR-10 and ImageNet, and show that the obtained sparse models can reach the state-of-the-art performance of dense models.
  arXiv  Detail & Related papers  (2020-06-12T15:07:08Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.